Comparison of processing speed of typical wavefront reconstruction methods for lateral shearing interferometry

Area of Science:

• Optics and Photonics
• Computational Science
• Metrology

Background:

• Lateral shearing interferometry is crucial for wavefront sensing and optical testing.
• Wavefront reconstruction speed and accuracy are key performance metrics.
• Comparing reconstruction methods like FFT, LSQ, and DZF is essential for optimizing interferometric systems.

Purpose of the Study:

• To design and evaluate optimized procedures for three wavefront reconstruction methods: iterative FFT (FFT), partial differential least-squares (LSQ), and difference Zernike polynomial fitting (DZF).
• To assess the calculation speeds of these methods on various GPUs and CPUs.
• To analyze the relationship between processing time and wavefront size across different hardware platforms.

Main Methods:

• Optimization of iterative FFT, LSQ, and DZF wavefront reconstruction algorithms.
• Benchmarking computational speeds using diverse Graphics Processing Units (GPUs) and Central Processing Units (CPUs).
• Analysis of processing time versus wavefront size (e.g., 1024x1024 pixels) on different hardware.

Main Results:

• The DZF method demonstrated the fastest performance, with processing times of 100 ms for a 1024x1024 wavefront.
• FFT and LSQ methods showed similar speeds on CPUs (449 ms and 494 ms, respectively), while FFT was faster on GPUs.
• CPUs outperformed GPUs for wavefront sizes smaller than 64x64 pixels, with less variation in performance between different CPUs compared to GPUs.

Conclusions:

• The DZF method provides a significant speed advantage for wavefront reconstruction in lateral shearing interferometry.
• Hardware choice (GPU vs. CPU) and wavefront size critically impact reconstruction speed.
• Optimized DZF method offers a superior solution for high-speed, accurate wavefront sensing applications.